2026-09-14

Beyond Solver Verdicts: Generative Reward Models for Autoformalization

👍 28

09/10 08:00

Neurosymbolic systems rely on mathematical solvers to guarantee reasoning correctness, yet solvers are fundamentally blind to whether a formal translation maintains strict reference-equivalence to a designated formalization. We formalize this vulnerability as Verdict-Preserving-Unfaithfulness (VPU):

Memory as Plans: World-Action Modeling with Memory-Grounded Planning

👍 36

09/10 08:00

Mainstream robotic policies often adopt a Markovian formulation, but many complex real-world manipulation tasks are inherently non-Markovian, requiring long-horizon memory beyond the current observation. Existing memory mechanisms often rely on language summaries, growing visual windows, or their co

Negative Self-Distillation: Learning to Reason by Avoiding Flaws

👍 29

09/10 08:00

On-Policy Self-Distillation (OPSD) has emerged as a popular paradigm for large language model (LLM) self-improvement, allowing models to act as their own teachers by leveraging privileged information such as ground-truth solutions. However, recent findings indicate that OPSD can severely degrade the

DRG-MAPPO: Hierarchical Dynamic Role-Graph Multi-Agent Reinforcement Learning for Cooperative Air Combat

👍 22

09/10 08:00

Multi-Agent Reinforcement Learning (MARL) has emerged as a pivotal paradigm for complex decision-making in autonomous systems and air combat. While MARL has demonstrated significant potential in air combat, achieving sophisticated tactical coordination remains a non-trivial challenge. This difficult

UniH^3: Unifying Hierarchical Homogeneity and Heterogeneity for All-in-One Medical Image Restoration

👍 24

09/10 08:00

All-in-One medical image restoration (MedIR) aims to address diverse tasks across modalities and degradation types using a single universal model. Existing methods typically prioritize modeling inter-task heterogeneity (e.g., distinct data distributions and degradation types). However, they largely

X-AuT: Progressive Audio-Encoder Compression for Speech LLMs with Cross-Scale Distillation

👍 41

09/10 08:00

Reducing audio-encoder depth lowers the inference cost of speech large language models, but removing complete blocks perturbs the embeddings consumed by the decoder and can cause deletion and premature end-of-sequence errors. We introduce X-AuT, a progressive framework that selects layer combination

Generative Late-Interaction Embeddings For Visual Document Retrieval

👍 25

09/10 08:00

Late-interaction retrieval is the state-of-the-art for visual document search, but it pays for its accuracy in storage. Existing compression methods retain a subset or local average of the N~1,000 vectors per page. Under aggressive storage budgets, however, these methods degrade sharply, and alterna

Recursive Code World Models: Building Complex Worlds through Recursive Scene Programs

👍 28

09/10 08:00

Code world models represent worlds as executable programs, but this representation alone does not determine how to construct a complex world. We introduce Recursive Code World Models (RCWM), a framework for reconstructing complex 3D worlds in code from a single reference image. RCWM couples a Recurs

T1: Terminal Agent Reinforcement Learning for Long-Horizon Tasks

👍 56

09/10 08:00

Agent usage is shifting toward long-horizon tasks such as coding and scientific discovery, among which terminal tasks are especially important. We introduce T1, a Mixture-of-Experts model of 122B total trained with reinforcement learning, operating a real shell in a cloud sandbox for up to 300+ tool

Think Before You Link: Rarity, Reasoning, and Retrieval in Multilingual Entity Linking

👍 22

09/09 08:00

Multimodal entity linking grounds entity mentions in text and images to knowledge-base entries. These systems degrade on rare entities, but prior work measures rarity primarily through popularity-based metrics such as pageviews. We broaden this view using knowledge-graph structural metrics that capt

Building Multilingual Bridges: Data Mixing as the Pillar of Generalization for In-Language Reasoning

👍 27

09/09 08:00

Reasoning language models have made substantial advances on a variety of complex tasks, yet their capabilities remain overwhelmingly English-centric: models primarily reason in English regardless of the language they are prompted in. This is inaccessible for non-English-speaking users, risks losing

IdeaAMBIG: Benchmarking Implementation-Critical Gaps in Research-Idea Specifications

👍 23

09/09 08:00

A research idea may be novel, coherent, and scientifically plausible, yet its proposed method may remain insufficiently specified for faithful implementation. We study the codification readiness of implementation-facing research-method specifications, defined by whether they provide sufficient metho

MetroLLM-Bench: Evaluating Language Models as Transit Kiosk Runtimes

👍 32

09/09 08:00

We introduce MetroLLM-Bench, a 955-case benchmark for testing language models as the policy layer of a transit kiosk. It covers six real metro systems, ranging from 37 to 414 stations, and eleven categories that include routing, fare calculation, disruptions, accessibility, and adversarial input. In

NCP-ArchPreview Technical Report: Moving towards Latent Space Language Models through Next Concept Prediction

👍 285

09/09 08:00

We introduce NCP-ArchPreview, a latent-space language model that pushes autoregressive pretraining beyond standard next-token prediction (NTP). Alongside NTP, the model learns through Next Concept Prediction (NCP) to predict discrete concepts that span multiple tokens, introducing an explicit and mo

The Semantic Bottleneck: Leveraging Semantic Representations for Non-Invasive Speech Decoding

👍 5

09/09 08:00

Non-invasive speech decoding remains constrained by the low signal-to-noise ratio of neural recordings, which makes fine-grained reconstruction of phonemes or individual words difficult. Motivated by neuroscientific evidence that high-level semantic representations are distributed across cortical re

ActReview: Rebuttal-Guided Training Data and Rubric Rewards for Actionable Peer Review Generation

👍 23

09/08 08:00

As LLMs are increasingly used for pre-submission self-review, there is growing demand for feedback that not only identifies weaknesses but also guides authors toward concrete revisions. We study this as Actionable Peer-review Generation and decompose it into two subtasks: diagnostic claim generation

SchemeArena: Factorized Stress Testing of Scheming in LLM Agents

👍 10

09/08 08:00

We study scheming in LLM agents, in which agents covertly pursue misaligned goals. Our focus is to understand how scheming arises from the interaction of key factors, such as instrumental goals, environmental affordances, oversight conditions, and perceived consequences. Prior work examines only a s

StochBench: A Domain-Specific Benchmark for Stochastic Processes in Lean

👍 9

09/08 08:00

Leading benchmarks for formal theorem proving with large language models are small collections drawn from competition math, such as the IMO and Putnam, that poorly represent field-specific applications. We introduce StochBench, a Lean 4 benchmark of 450 graduate stochastic-processes problems at vary

AgentGrad: Intervention-guided Prompt Optimization for Multi Agent Systems

👍 93

09/08 08:00

Large language model (LLM)-based multi-agent systems (MAS) achieve strong performance by employing specialized multiple agents, yet their performance depends on the prompt design of each agent. For MAS prompt optimization, textual gradient methods that guide prompt updates using natural-language fee

PlannerForge: LLM Agents for Scenario-Based Testing of Motion Planners in Autonomous Driving

👍 6

09/08 08:00

Ensuring the safety of autonomous driving is a critical challenge. Scenario-based testing is a systematic process used to validate Autonomous Driving Systems (ADSs), but it remains a fragmented modular pipeline in which scenario generation, retrieval, modification, ADS execution, and results analysi

SpatialBlock: Enhancing Spatial Intelligence in LVLMs via Synthetic Block-Stacking Problem

👍 124

09/07 08:00

Large Vision-Language Models (LVLMs) have achieved strong performance on diverse visual tasks, yet their ability to reconstruct and reason about the 3D structure of the scene depicted in 2D images -- referred to as spatial intelligence -- remains limited. Existing approaches attempt to address this

CARDEA: Auditable Reasoning Grounded in Spatial Evidence for End-to-End Coronary Angiography Interpretation

👍 24

09/07 08:00

Invasive coronary angiography (CAG) is the gold standard for diagnosing coronary artery disease, but interpretation varies substantially among observers. Existing AI systems can improve consistency but lack auditable decision processes and are limited in comprehensive open-ended assessment, undermin

PARSER: Read in Parallel, Reason in Depth for Long-Context LLM Agents

👍 23

09/06 08:00

Sequential memory agents process long documents by reading chunks one after another while maintaining a compact memory state, coupling document traversal to reasoning depth. This coupling introduces sensitivity to evidence placement and ties inference latency linearly to document length. We introduc

OracleZoom: On-Policy Self-Distillation Inspired Reference-Constrained Recursive Image Super Resolution

👍 13

09/06 08:00

Recursive Super-Resolution (SR) extends fixed-scale SR to extreme magnification by repeatedly feeding predictions back into the same model, analogous to zooming an image repeatedly. However, ground truth availability at every scale, especially at depth, remains challenging as the required source res

EvoSafeHarness: Evolving Model- and Domain-Specific Harnesses for Securing Agents

👍 59

09/05 08:00

Large Language Model (LLM) agents are turning language into real-world effects, making safety necessary against both indirect prompt injections and direct harmful requests. System-level safety harnesses add an enforcement layer beyond model-level defenses, but existing harnesses are usually designed

From Reweighting to Rewriting: Unlocking the Intervention Effects of Influential Samples in Training Data Attribution

👍 3

09/02 08:00

Training data attribution (TDA) aims to identify training examples that shape model behavior, but its intervention value depends on both which examples are selected and how they are modified. Influence functions (IF) estimate behavioral changes under infinitesimal reweighting, yet IF-selected exampl

TempCloze: Can Video-LLMs Identify the Missing Middle?

👍 29

09/01 08:00

Temporal reasoning benchmarks for Video-LLMs are often mediated by language, leaving room for linguistic shortcuts from option wording, answer correlations, or language priors. To reduce such shortcuts, we introduce TempCloze, a video cloze benchmark for evaluating visual temporal reasoning in Video

Scaling Automatic Research Agents via World Models

👍 447

08/29 08:00

Automating empirical research is a long-standing direction of AI. Recent automatic research (AutoResearch) agents bring this goal within reach, as modern LLMs show the capability to independently implement solutions and learn from the execution outcomes. Behind these gains, post-training (especially

HyQuant: Hybrid-Precision Quantization for LLM Attention

👍 30

08/28 08:00

Quantization has been widely adopted in LLM training and inference to reduce cost and improve efficiency. However, low-bit quantization of the attention module often introduces large errors at very low bit-widths, causing performance degradation. Existing methods mainly rely on smoothing techniques

A Three-Layer Caching Architecture for Low-Latency LLM Web Search on Commodity CPU Hardware

👍 11

08/12 08:00

AI-powered search products such as ChatGPT search, Google's AI Overviews, and Perplexity provide LLM-synthesized answers grounded in live web results. We developed OreoLook (formerly lixSearch), an open-source answer engine using automated browser agents and provider-routed LLM inference. Its local

What if AGI is already here, and it's made of... children?

@IAmPascio · 124.8K 粉丝 · 812.5K 阅 · 881 赞 · 140 转

09/12 23:41

The superintelligence, the singularity, AGI, whatever you call it, the one we're so scared of is, ironically, also the one we most likely won't see coming. This essay is going to probably be a pretty

中文介绍 探讨超级智能（AGI）的来临及其表现形式，引发对人类未来的思考。

How to build a full game with Blender and GPT 5.6 Astra (not just a demo)

@builtbysketch · 1.4K 粉丝 · 717.2K 阅 · 518 赞 · 79 转

09/12 22:00

I built a full game with GPT-6 Astra, and it kinda blew up on X. Initially it was just the demo, but 4 days later we have the full playable Okay, not just a crappy one-run demo. Check PaperRoute

中文介绍 介绍使用 Blender 和 GPT-6 Astra 开发完整游戏的经验，从演示到完整可玩版本。

No, OpenAI Did Not “Solve Navier–Stokes.”

@mayukh_panja · 9.8K 粉丝 · 311.6K 阅 · 501 赞 · 75 转

09/11 21:44

There seems to be a widespread impression that OpenAI has somehow “solved the Navier–Stokes equations,” and that this represents some enormous breakthrough for fluid dynamics and engineering. It

中文介绍 澄清 OpenAI 没有解决 Navier-Stokes 方程，探讨该误解的来源及其影响。

Intelligence Has a Speed Limit

@docmilanfar · 116.7K 粉丝 · 126.0K 阅 · 512 赞 · 63 转

09/13 05:32

In AI circles, Recursive Self-Improvement (RSI) is talked about in almost magical terms: a model tweaks itself, becomes smarter, uses that intellect to rewrite itself again, and this leads to

中文介绍 分析 AI 递归自我改进（RSI）的速度限制及其在 AI 领域的应用。

A Severe Misalignment of AI and Cancer Biology

@mbeisen · 73.0K 粉丝 · 113.7K 阅 · 528 赞 · 68 转

09/12 09:19

Over the last few months, the cancer-curing capabilities of LLMs have improved dramatically, to the point that they have cured previously untreatable types of cancer and are on the cusp of developing

中文介绍 探讨 LLM 在癌症生物学中的误用，以及其可能带来的严重后果。

How to automate content creation (Full Course)

@EXM7777 · 142.6K 粉丝 · 94.6K 阅 · 501 赞 · 56 转

09/11 22:13

I'm basically giving you the full system i'd use to automate content creation for ANY business... every tool in the chain, every step in order, and where it breaks because a content agent is just a

中文介绍 分享自动化内容创建的全套系统，包括工具链和每个步骤的细节。

How to start vibe editing (No Fluff Guide)

@ForwardEditor · 11.7K 粉丝 · 57.6K 阅 · 511 赞 · 32 转

09/11 11:53

Hey, I’m Brian. Let’s get you started with vibe editing. 1. What I highly suggest having DaVinci Resolve Studio - $295 USD. It’s a one-time purchase, not a subscription. In my experience, Studio

中文介绍 介绍 vibe editing 的入门指南，推荐使用 DaVinci Resolve Studio。

GPT-6 Astra by our community

@OpenAIDevs · 421.3K 粉丝 · 51.6K 阅 · 677 赞 · 54 转

09/13 01:33

We shipped GPT-6 Astra. Builders took it from there: Explore 2,234 modeled anatomical parts in 3D: A Manhattan recreation built in Unreal Engine: Six Van Gogh paintings become a walkable

中文介绍 展示 GPT-6 Astra 的多功能应用，包括 3D 解剖结构、曼哈顿重建和梵高画作的可行走模型。

AI Engineering Skills Map: Shaping the build

@AndrewYNg · 1.9M 粉丝 · 43.5K 阅 · 722 赞 · 105 转

09/12 01:11

When you’re skilled at AI Engineering, your best work won’t be merely implementing a product that someone else spec’ed out. Instead, you will actively shape the build. Before modern AI tools

中文介绍 提出 AI 工程技能图谱，强调在产品构建中的主动塑造作用。

The Idiot Boss approach to getting more Astra/Fable quota

@anshuc · 15.6K 粉丝 · 41.4K 阅 · 544 赞 · 36 转

09/13 00:31

Most people say to use the big, expensive model as an orchestrator, delegating grunt work to small, cheap workers. That's the opposite of what you should do. Here's why. Comparing different approaches

中文介绍 提出“傻瓜老板”策略，以获得更多 Astra/Fable 配额，并对比不同方法。

The Model Is Becoming Cheaper. The Context Is Not.

@Didotxyz_ · 63.6K 粉丝 · 17.9K 阅 · 516 赞 · 388 转

09/12 22:46

Model access is becoming easier to buy, compare, route, and replace. Public API menus now include premium reasoning models, lower-cost variants, cached inputs, batch processing, and models optimized

中文介绍 讨论模型价格下降与使用环境复杂性增加的对比，强调实际应用中的挑战。

Perplexity trusts GPT-6 Astra with end-to-end systems

OpenAI News · 09/14 08:00

Perplexity uses Astra to write communications, change software, and monitor production systems, and checks in much less frequently than with earlier models.

中文介绍 Perplexity采用Astra进行通信写作、软件更改和生产系统监控，比早期模型检查频率低得多。

The Rise of the Forward Deployed Engineer — and How To Do the Job Right

Latent Space · 09/12 23:01

Before co-founding Kepler, Vinoo Ganesh led Spark at Palantir and built Project Frontline — a pioneering program for Forward Deployed Engineers. He takes us through the best practices of FDEs.

中文介绍 Vinoo Ganesh介绍了FDE的最佳实践，他曾是Palantir的Spark领导者和Project Frontline的创建者。

[AINews] DeepSeek v4.1-Flash: 763B-P8B-D16B novel causal Encoder–Decoder architecture with vision marks the Return of the Whale

Latent Space · 09/12 13:56

We agree with Sebastian: this should have been DeepSeek v5

中文介绍 DeepSeek v4.1-Flash引入了763B-P8B-D16B新型因果编码器-解码器架构，标志着鲸鱼的回归。

Roundtables: Could AI really kill us all?

MIT Tech Review AI · 09/12 04:05

Employees at the world’s leading AI labs are saying there’s a real possibility that advanced AI could destroy humanity. Are they right? Or is this more scaremongering and hype? Join MIT Technology Review executive editor Niall Firth for a conversation with senior AI editor Will Douglas Heaven and AI

中文介绍 全球领先AI实验室的员工表示，高级AI有可能毁灭人类。

Cognition helps Devin test its own work with GPT‑6 Astra

OpenAI News · 09/12 00:00

GPT‑6 Astra improves Devin’s ability to test software and show that it works, with the goal of helping engineers review less code and ship more.

中文介绍 GPT-6 Astra帮助Devin测试其工作，以提高测试软件和证明其工作能力，目标是帮助工程师审查更少的代码并更快地发布。

Rapidly scaling online storage to serve over 1 billion ChatGPT users

OpenAI News · 09/11 18:00

Learn how OpenAI evolved Habitat from a Python library into a globally distributed storage platform serving 1 billion ChatGPT users and 22M requests per second.

中文介绍 OpenAI将Habitat从Python库演变成全球分布式存储平台，为10亿ChatGPT用户和每秒2200万请求提供服务。

Agents API 🤖, Cognition SWE-2 👨‍💻, Muse Shared Agents 🧑‍🤝‍🧑

TLDR AI · 09/11 08:00

中文介绍 TLDR AI发布了一系列AI相关更新。

How a researcher uses Codex and ChatGPT to search for new antimicrobial molecules

OpenAI News · 09/11 00:00

César de la Fuente’s lab uses Codex and ChatGPT to search living and extinct genomes for antimicrobial candidates to fight drug-resistant infections.

中文介绍 César de la Fuente的实验室使用Codex和ChatGPT搜索活性和灭绝生物的基因组，寻找对抗耐药性感染的抗菌候选分子。

Now everyone can put data to work

OpenAI News · 09/10 23:00

Meet the Data agent in ChatGPT Work. Connect company data, uncover insights, and build interactive dashboards with AI using natural language.

中文介绍 ChatGPT Work中的数据代理可连接公司数据，发现洞察力，并使用AI构建交互式仪表板。

Powering AI is an architecture problem

MIT Tech Review AI · 09/10 19:00

On July 22, 2026, a transmission line fault in Ashburn, Virginia—the heart of the world’s largest data center cluster—knocked more than 3 gigawatts of load off the grid in seconds. And it wasn’t the first time. Two years earlier, a single failed surge arrester dropped roughly 60 Virginia facilities

中文介绍 2026年7月22日，弗吉尼亚州阿什本——全球最大数据中心集群的心脏地带发生输电线路故障，几秒钟内从电网中切除了30多吉瓦的负荷。

Introducing ChatGPT for Financial Services

OpenAI News · 09/10 15:00

Introducing ChatGPT for Financial Services, combining built-in financial data and GPT-6 Astra for research, modeling, and client-ready materials.

中文介绍 OpenAI推出ChatGPT金融服务平台，结合内置金融数据和GPT-6 Astra进行研究和建模。

Expanding AI access and cyber defense for federal, state, local, and tribal governments

OpenAI News · 09/10 15:00

OpenAI and GSA will offer eligible federal, state, local, and tribal governments $0 license fees, 50% off usage, and expanded cyber defense support.

中文介绍 OpenAI和GSA将为符合条件的联邦、州、地方和部落政府提供0美元许可费、50%的使用折扣和扩展的网络安全支持。

[AINews] not much happened today

Latent Space · 09/10 11:33

a quiet day

中文介绍 今天是一个平静的日子。

Build more natural voice experiences with GPT‑Live‑1 in the API

OpenAI News · 09/10 08:00

GPT‑Live‑1 brings natural, full-duplex voice conversations to the API, with stronger instruction following, custom voices, and telephony support.

中文介绍 GPT-Live-1将自然、全双工语音对话引入API，具有更强的指令遵循、定制声音和电话支持。

Introducing the Agents API

OpenAI News · 09/10 08:00

Build and launch cloud agents with the Agents API, a managed service powered by the Codex harness for orchestration, long-running sessions, and tool use.

中文介绍 使用Agents API构建和部署云代理，该API是一个由Codex驱动器支持的托管服务，用于编排、长时间会话和工具使用。